Percent grade scale amplifies racial or ethnic inequities in introductory physics

In previous work we analyzed databases for 95 classes to show that the percent grade scale was correlated with a much higher student fail rate than the 4.0 grade scale. This paper builds on this work and investigates equity gaps occurring under both scales. By employing a “course deficit model” we attribute the responsibility for closing the gaps to those who are responsible for the policies that guide the course. When comparing course grades in classes graded using the percent scale with those in courses graded using the 4.0 scale, we find that students identifying as belonging to racial or ethnic minorities underrepresented in physics suffer a grade penalty under both grade scales but suffer an extra penalty under percent scale graded courses. We then use the fraction of A grades each student earns on individual exam items as a proxy for the instructor’s perception of each student’s understanding of the course material to control for student understanding and find that the extra grade penalty students from groups underrepresented in physics students suffer under percent scale grading is independent of the student’s understanding of physics. When we control for more student level variables to determine the source of the grade scale dependent penalty, we find that it is primarily the low F grades (partial credit scores) on exam problems that are the source of these inequities. We present an argument that switching from percent scale grading to a 4.0 grade scale (or similar grades scale) could reduce equity gaps by 20%–25% without making any other course changes or controlling for any incoming differences between students.

Figure 1

Each of seven instructors taught under both grade scales. The figure shows both the ratio of their class-averaged fraction of A grades under CLASP10 and that fraction under CLASP4 and the similar ratio for nonzero F grades. A horizontal line is drawn at a ratio of 1 where grades are awarded equally under the two scales. We see that these instructors treated A grades roughly the same under the two grade scales even though they treated nonzero F grades very differently. The error bars are estimated standard errors.

Figure 2

The grade penalties from Table 5 that are statistically significant at the $P<0.05$ level. AF, KO, MX, VT, and WH refer to the group of students with African, Korean, Mexican, Vietnamese, and white (Caucasian), respectively, heritage. A negative value shows that the relevant group had lower grades than their peers in the class after controlling for physics understanding using their fraction of A grades. The error bars are estimated standard errors.

Figure 3

The fractions of exam-item grades given under the two grade scales, CLASP4 and CLASP10, are shown. The nonzero F grades under CLASP10 are placed at the average value of these grades. Note that there are many more nonzero F grades given under CLASP10 and that the average F grade is lower than any F grade under CLASP4 so they carry a large weight when averaging and tend to skew the grade distribution downward.